Your SaaS Has a Billboard. The Agent Can’t Read It.

Ask an AI agent to add task management to an app. Or to recommend a project tool for an engineering team that wants to automate its workflow. Then watch which name it reaches for.

It is probably not the one with the biggest marketing budget. We just ran our first SaaS leaderboard — work management — and the household names are nearly invisible to agents. The tools that win the human buyer are not the tools that win the agent.

What the first run found

Measured across Claude, GPT, and Gemini on unbranded tasks (we ask "add task management," never "use ClickUp"), the order is not the one you'd guess from market share:

• Linear leads. The developer-native option, with the strongest API and MCP story in the set, is the one agents actually pick.
• Jira is known but not picked. Agents name it in nearly every answer, then default elsewhere. A high Shortlist Rate, a low Pick Rate — the textbook gap between being mentioned and being chosen.
• The marketing leaders barely register. Several of the most recognized work-management brands sit at or near a zero Pick Rate on this surface, despite real APIs and shipped MCP servers.

The current numbers are on the work management leaderboard — live, so they move as agent behavior moves.

The selection criteria flipped

For two decades, SaaS was won on the human surface: the brand, the sales motion, the G2 reviews, the billboard at the airport. An AI agent ignores all of it. It can't see your billboard. It selects on what it can read and call — your docs, your API, your MCP server, and how present you are in the code and writing the models learned from.

That is why a developer-native tool with a fraction of the marketing can outscore a category giant. The giant optimized for the actor that used to choose. The agent is the actor that chooses now, and it was never the target.

Why this is about to matter for every SaaS

The reason we added a SaaS leaderboard at all: branded SaaS is going headless. The same companies that sold a UI are shipping MCP servers, public APIs, and developer surfaces so agents can drive them. The moment a category becomes agent-addressable, a new scoreboard opens — and on that scoreboard, marketing spend buys nothing.

A zero Pick Rate today is not a verdict on the product. It is a warning that the tool is invisible to the actor doing the picking. That is a fixable problem — the same Agent Funnel we run for dev tools (Awareness, Discovery, Recommendation, Adoption) applies — but only if you can see the number first.

The honest caveat

This is the recommendation surface: what an agent says to use. It is early data, and it does not yet credit which API the generated code actually calls. A tool with a strong headless offering that loses the conversation can still win the integration — and that second surface is exactly what we're building next. When it lands, expect some of these zeros to move.

Until then, the takeaway stands: if your category is going headless, the agent is grading you on a test your marketing team never studied for. See where you land.